org.grouplens.lenskit.eval.metrics

Interface Metric<E>

Type Parameters:

E - The type of evaluation.

All Known Subinterfaces:

ModelMetric, TestUserMetric

All Known Implementing Classes:

AbstractMetric, AbstractTestUserMetric, CoveragePredictMetric, EntropyPredictMetric, FunctionModelMetric, FunctionMultiModelMetric, HLUtilityPredictMetric, IndependentRecallTopNMetric, MAEPredictMetric, NDCGPredictMetric, NDCGTopNMetric, PrecisionRecallTopNMetric, RMSEPredictMetric, TopNEntropyMetric, TopNLengthMetric, TopNPopularityMetric

public interface Metric<E>

Base interface for metrics, which are used by evaluations to measure their results.

Since:

0.10

Method Summary

Methods

Modifier and Type	Method and Description
void	finishEvaluation() Finish the evaluation, releasing any resources allocated for it.
void	startEvaluation(E eval) Initialize the metric to accumulate evaluations for the specified evaluation.

Method Detail

startEvaluation

void startEvaluation(E eval)

Initialize the metric to accumulate evaluations for the specified evaluation. This method is called before any accumulators are constructed. It is invalid to start a metric twice without finishing it first.

Parameters:

eval - The evaluation this metric is in use for.

finishEvaluation

void finishEvaluation()

Finish the evaluation, releasing any resources allocated for it. This will be called after all algorithms and data sets are run through the accumulator. After this method, the metric should be ready to be used on another evaluation.

Review:

Is that really the behavior we want?